When master_use_gtid=no, the IO thread loads the slave GTID state from
the master during connect. This races with the SQL thread when
gtid_ignore_duplicates=1. If an event is in the relay log from before
the new connect and has not been applied yet, moving the slave
position causes the SQL thread to think that event should be skipped
due to gtid_ignore_duplicates=1.
This patch simply disables gtid_ignore_duplicates when not using GTID,
which seems to be what one would expect.
The following is an updated commit message for the following commit
that was pushed before I had a chance to update the commit message:
c5e25c8b40
Fixed dead locks when doing stop slave while slave was starting.
- Added a separate lock for protecting start/stop/reset of a specific slave.
This solves some possible dead locks when one calls stop slave while
the slave is starting as the old run_locks was over used for other things.
- Set hash->records to 0 before calling free of all hash elements.
This was set to stop concurrent threads to loop over hash elements and
access members that was already freed.
This was a problem especially in start_all_slaves/stop_all_slaves
as the mutex protecting the hash was temporarily released while a slave
was started/stopped.
- Because of change to hash->records during hash_reset(),
any_slave_sql_running() will return 1 during shutdown as one can't
loop over master_info_index->master_info_hash while hash_reset() of it
is in progress.
This also fixes a potential old bug in any_slave_sql_running() where
during shutdown and ~Master_info_index(), my_hash_free() we could
potentially try to access elements that was already freed.
The problem was that waiting for pause_for_ftwrl was done before
event_group was completed. This caused rpl_pause_for_ftwrl() to wait
forever during FLUSH TABLES WITH READ LOCK.
Now we only wait for FLUSH TABLES WITH READ LOCK when we are changing
to a new event group.
The reason for this is that stop slave takes LOCK_active_mi over the
whole operation while some slave operations will also need LOCK_active_mi
which causes deadlocks.
Fixed by introducing object counting for Master_info and not taking
LOCK_active_mi over stop slave or even stop_all_slaves()
Another benefit of this approach is that it allows:
- Multiple threads can run SHOW SLAVE STATUS at the same time
- START/STOP/RESET/SLAVE STATUS on a slave will not block other slaves
- Simpler interface for handling get_master_info()
- Added some missing unlock of 'log_lock' in error condtions
- Moved rpl_parallel_inactivate_pool(&global_rpl_thread_pool) to end
of stop_slave() to not have to use LOCK_active_mi inside
terminate_slave_threads()
- Changed argument for remove_master_info() to Master_info, as we always
have this available
- Fixed core dump when doing FLUSH TABLES WITH READ LOCK and parallel
replication. Problem was that waiting for pause_for_ftwrl was not done
when deleting rpt->current_owner after a force_abort.
The SQL thread keeps track of the position in the current relay log from
which to read the next event. This position is not normally used, but a
certain interaction with the IO thread can cause the SQL thread to re-open
the relay log and seek to the stored position.
In parallel replication, there were a couple of places where the position
was not updated. This created a race where a re-open of the relay log could
seek to the wrong position and start re-reading and processing events
already handled once, causing various kinds of problems.
Fix this by moving the position update into a single place in
apply_event_and_update_pos(), which should ensure that the position is
always updated in the parallel replication case.
This problem was found from the testcase of MDEV-10863, but it is logically
a separate problem.
The function apply_event_and_update_pos() is called with the
rli->data_lock mutex held. However, there seems to be nothing in the
function actually needing the mutex to be held. Certainly not in the
parallel replication case, where sql_slave_skip_counter is always 0
since the non-zero case is handled by the SQL driver thread.
So this patch makes parallel replication use a variant of
apply_event_and_update_pos() without the need to take the
rli->data_lock mutex. This avoids one contended global mutex for each
event executed, which might improve performance on CPU-bound workloads
somewhat.
Signed-off-by: Kristian Nielsen <knielsen@knielsen-hq.org>
When a deadlock kill is detected inside the storage engine, the kill
is not done immediately, to avoid calling back into the storage engine
kill_query method with various lock subsystem mutexes held. Instead the
kill is queued and done later by a slave background thread.
This patch in preparation for fixing TokuDB optimistic parallel
replication, as well as for removing locking hacks in InnoDB/XtraDB in
10.2.
Signed-off-by: Kristian Nielsen <knielsen at knielsen-hq.org>
This includes fixing all utilities to not have any memory leaks,
as safemalloc warnings stopped tests from passing on MacOSX.
- Ensure that all clients takes character-set-dir, as the
libmysqlclient library will use it.
- mysql-test-run now passes character-set-dir to all external clients.
- Changed dynstr_free() so that it can be called twice (made freeing code easier)
- Changed rpl_global_gtid_slave_state to be allocated dynamicly as it
includes a mutex that needs to be initizlied/destroyed before my_end() is called.
- Removed rpl_slave_state::init() and rpl_slave_stage::deinit() as
their job are better handling by constructor and delete.
- Print alias instead of table_name in check_duplicate_key as
table_name may have been converted to lower case.
Other things:
- Fixed a case in time_to_datetime_with_warn() where we where
using && instead of & in tests
Problem was that we used same condition variable with 2 different mutex.
Fixed by changing to use COND_rpl_thread_stop instead of COND_parallel_entry
for stopping threads.
Patch by Kristian Nielsen
Problem is that FLUSH TABLES WITH READ LOCK first blocks threads from
starting new commits, then waits for running commits to complete. But
in-order parallel replication needs commits to happen in a particular
order, so this can easily deadlock.
To fix this problem, this patch introduces a way to temporarily pause
the parallel replication worker threads. Before starting FTWRL, we let
all worker threads complete in-progress transactions, and then
wait. Then we proceed to take the global read lock. Once the lock is
obtained, we unpause the worker threads. Now commits are blocked from
starting by the global read lock, so the deadlock will no longer occur.
Before, the Seconds_behind_master was updated already when an event
was queued for a worker thread to execute later. This might lead users
to interpret a low value as the slave being almost up to date with the
master, while in reality there might still be lots and lots of events
still queued up waiting to be applied by the slave.
See https://lists.launchpad.net/maria-developers/msg08958.html for
more detailed discussions.
The code was using the wrong variable when comparing the binlog name
for the UNTIL position. This could cause the comparison to fail after
binlog rotation, in turn causing the UNTIL clause to not trigger slave
stop.
The assertion is there to catch cases where we rollback while
mark_start_commit() is active. This can allow following event groups
to be replicated too early, causing conflicts.
But in this case, we have an _explicit_ ROLLBACK event in the binlog,
which should not assert.
We fix this by delaying the mark_start_commit() in the explicit
ROLLBACK case. It seems safest to delay this in ROLLBACK case anyway,
and there should be no reason to try to optimise this corner case.
This bug is essentially another variant of MDEV-7458.
If a transaction conflict caused a deadlock kill of T2 in record_gtid()
during commit, the code would do a rollback _before_ running
rgi->unmark_start_commit(). This creates a race where following transactions
could start too early (before T2 has completed its transaction retry). This
in turn could lead to replication failure, if there was a conflict that
caused eg. duplicate key error or similar.
The fix is to remove these rollbacks (in Query_log_event::do_apply_event()
and Xid_log_event::do_apply_event(). They seem out-of-place; code in
log_event.cc generally does not roll back on error, this is handled higher
up.
In addition, because of the extreme difficulty of reproducing bugs like
MDEV-7458 and MDEV-8302, this patch adds some extra precations to try to
detect (in debug builds) or prevent (in release builds) similar bugs.
ha_rollback_trans() will now call unmark_start_commit() if needed (and
assert in debug build when a caller does rollback without unmark first).
We also add an extra check for thd->killed() so that we avoid doing
mark_start_commit() if we already have a pending deadlock kill.
And we add a missing unmark_start_commit() call in the error case, found by
the above assertion.
When the slave processes the master restart format_description event,
parallel replication needs to complete any prior events before processing
the restart event (which closes temporary tables and such stuff).
This happens in wait_for_workers_idle(), however it was not waiting long
enough. The wait was using wait_for_prior_commit(), but at that points table
can still be open. This lead to assertion in this case.
So change wait_for_workers_idle() to wait until all worker threads have
reached finish_event_group(), at which point all tables should have been
closed.
In optimistic parallel replication, it is not safe to try to run a following
transaction in parallel with a DDL statement, and there is code to prevent
this.
However, the code was missing the case where the DDL is the very first event
after slave start. In this case, following transactions could run in
parallel with the DDL, which can cause the slave to hang or even corrupt
slave in unlucky cases.
There was a rare race, where a deadlock error might not be correctly
handled, causing the slave to stop with something like this in the error
log:
150423 14:04:10 [ERROR] Slave SQL: Connection was killed, Gtid 0-1-2, Internal MariaDB error code: 1927
150423 14:04:10 [Warning] Slave: Connection was killed Error_code: 1927
150423 14:04:10 [Warning] Slave: Deadlock found when trying to get lock; try restarting transaction Error_code: 1213
150423 14:04:10 [Warning] Slave: Connection was killed Error_code: 1927
150423 14:04:10 [Warning] Slave: Connection was killed Error_code: 1927
150423 14:04:10 [ERROR] Error running query, slave SQL thread aborted. Fix the problem, and restart the slave SQL thread with "SLAVE START". We stopped at log 'master-bin.000001 position 1234
The problem was incorrect error handling. When a deadlock is detected, it
causes a KILL CONNECTION on the offending thread. This error is then later
converted to a deadlock error, and the transaction is retried.
However, the deadlock error was not cleared at the start of the retry, nor
was the lingering kill signal. So it was possible to get another deadlock
kill early during retry. If this happened with particular thread
scheduling/timing, it was possible that the new KILL CONNECTION error was
masked by the earlier deadlock error, so that the second kill was not
properly converted into a deadlock error and retry.
This patch adds code that clears the old error and killed flag before
starting the retry. It also adds code to handle a deadlock kill caught in a
couple of places where it was not handled before.
The hangs occur when the group_commit_orderer object is freed before the last
mark_start_commit() call on it - this loses the wakeup to other waiting worker
threads, causing them to hang until killed manually.
The object was freed because wakeup_subsequent_commits() was called two early
in two places. For MDEV-7888, during ANALYZE TABLE, and for MDEV-7929 during
record_gtid() after processing a DDL event. The group_commit_orderer object
can be freed when its last transaction has called wait_for_prior_commit().
Fix by implementing a suspend/resume mechanism for wakeup_subsequent_commits()
that can be used in places where a transaction is committed without this being
the commit of the actual replication event group.
Also add a protection mechanism (that asserts in debug builds) which can
prevent the too-early free and hang if other similar bugs should remain in
other parts of the code.
This patch fixes a bug in the error handling in parallel replication, when one
worker thread gets a failure and other worker threads processing later
transactions have to rollback and abort.
The problem was with the lifetime of group_commit_orderer objects (GCOs).
A GCO is freed when we register that its last event group has committed. This
relies on register_wait_for_prior_commit() and wait_for_prior_commit() to
ensure that the fact that T2 has committed implies that any earlier T1 has
also committed, and can thus no longer execute mark_start_commit().
However, in the error case, the code was skipping the
register_wait_for_prior_commit() and wait_for_prior_commit() calls. Thus
commit ordering was not guaranteed, and a GCO could be freed too early. Then a
later mark_start_commit() would reference deallocated GCO, which could lead to
lost wakeup (causing slave threads to hang) or other corruption.
This patch makes also the error case respect commit order. This way, also the
error case gets the GCO lifetime correct, and the hang no longer occurs.
When a transaction in parallel replication needs to retry (eg. because of
deadlock kill), first wait for all prior transactions to commit before doing
the retry. This way, we avoid the retry once again conflicting with a prior
transaction, requiring yet another retry.
Without this patch, we saw "in the wild" that transactions had to be retried
more than 10 times to succeed, which exceeds the default
--slave_transaction_retries value and is in any case undesirable.
(We already do this in 10.1 in "optimistic" parallel replication mode; this
patch just makes the code use the same logic for "conservative" mode (only
mode in 10.0)).
The patch for optimistic parallel replication as a memory optimisation moved
the gco->installed field into a bit in gco->flags. However, that is just plain
wrong. The gco->flags field is owned by the SQL driver thread, but
gco->installed is used by the worker threads, so this will cause a race
condition.
The user-visible problem might be conflicts between transactions and/or slave
threads hanging.
So revert this part of the optimistic parallel replication patch, going back
to using a separate field gco->installed like in 10.0.
Delay spawning parallel replication worker threads until a slave SQL
thread is running, and de-spawn them when the last SQL thread stops.
This is especially useful to avoid needless threads on a master in a
setup where same my.cnf is used on masters and slaves.
The problem occurs in parallel replication in GTID mode, when we are using
multiple replication domains. In this case, if the SQL thread stops, the
slave GTID position may refer to a different point in the relay log for each
domain.
The bug was that when the SQL thread was stopped and restarted (but the IO
thread was kept running), the SQL thread would resume applying the relay log
from the point of the most advanced replication domain, silently skipping all
earlier events within other domains. This caused replication corruption.
This patch solves the problem by storing, when the SQL thread stops with
multiple parallel replication domains active, the current GTID
position. Additionally, the current position in the relay logs is moved back
to a point known to be earlier than the current position of any replication
domain. Then when the SQL thread restarts from the earlier position, GTIDs
encountered are compared against the stored GTID position. Any GTID that was
already applied before the stop is skipped to avoid duplicate apply.
This patch should have no effect if multi-domain GTID parallel replication is
not used. Similarly, if both SQL and IO thread are stopped and restarted, the
patch has no effect, as in this case the existing relay logs are removed and
re-fetched from the master at the current global @@gtid_slave_pos.
If somehow the COMMIT or XID event in an event group was missing, the code in
parallel replication to handle this was not sufficient, leading to server
deadlock.
